Segmentation of blood cells is a difficult task because of the presence of noise and the substantial brightness changes within cells in microscopy images. A method based on the Hough transform and fuzzy curve tracing is proposed in this paper. The Hough transform is used to detect the rough circular boundary of each cell. Then, fuzzy curve tracing is employed to detect the exact cell boundary. This approach reduces the effects of noise and the uneven brightness within the cells effectively. In addition, it can even separate slightly overlapping cells. Experiment results show that the proposed method is superior to many existing segmentation methods.